text model · SmolLM2 · iOS
Can I run SmolLM2 135M on iPhone 16 Pro?
Yes. SmolLM2 135M runs on iPhone 16 Pro at Q4_K_M (~1 GB of ~4.5 GB usable).
Runs at Q4_K_M using ~1 GB of ~4.5 GB usable. You have room for FP16 for higher quality.
That figure is at a 4k context and moves about ±15% as context length changes. iPhone 16 Pro leaves ~3.5 GB of headroom, room to step up to FP16 for higher quality.
- Q4_K_M needed
- ~1 GB
- Usable on device
- ~4.5 GB
- Device memory
- 8 GB
- Best quant
- Q4_K_M
Which quant fits
Running cost · estimate
- Power draw
- ~12 W
- Electricity / 1M tokens
- ~$0
- Pays for itself after
- ~1,998M tok
At ~$0.15/kWh and the estimated ~286 tok/s, a million generated tokens costs about $0 in electricity. Versus a hosted API at ~$0.5 per million tokens, the ~$999 iPhone 16 Pro pays for itself after roughly 1,998 million tokens, so local hardware is mostly a fixed cost, not a per-token one. TDP is the peak draw, so this is an upper bound. Assumptions.
How to run it
On iOS use Apple Foundation Models (Built into iOS 26, ~3B on-device model, zero download, fully private.).
- Parameters
- 0.135B
- Q4_K_M size
- 0.105 GB
- Q8_0 size
- 0.145 GB
- Context
- 2k
- Ollama tag
- smollm2:135m
- Memory
- 8 GB unified
- Usable for weights
- ~4.5 GB
- Power draw
- ~12 W
- Best runtime
- llama.cpp + Metal (via PocketPal or Off Grid app)
You could also run
Run SmolLM2 135M on other hardware
FAQ
Can iPhone 16 Pro run SmolLM2 135M?
Yes. SmolLM2 135M runs on iPhone 16 Pro at Q4_K_M (~1 GB of ~4.5 GB usable).
How much memory does SmolLM2 135M need?
iPhone 16 Pro has room to spare. At Q4_K_M the weights are ~0.105 GB; with KV cache and runtime overhead, budget ~1 GB at a 4k context.
What is the best tool to run SmolLM2 135M on iOS?
On iPhone and iPad, Apple Foundation Models (Built into iOS 26, ~3B on-device model, zero download, fully private.) is the standard choice. Phones realistically run 1B-4B class models. Anything larger thermally throttles or OOMs.
Embed this
A live badge for your model card or README, updated as the data is. [](https://localmodel.run/can-i-run/smollm2-135m/iphone-16-pro) Sources
Weights are measured from GGUF files; KV cache and overhead are computed, so totals can vary ~15% with context and runtime. Any tok/s is a bandwidth estimate. See methodology.